python-overfitting

Multi-LCB extends the LiveCodeBench benchmark to evaluate LLMs across twelve programming languages while preserving contamination controls, revealing Python overfitting and language-specific contamination issues.